Skip to content

Use computer tools that let students explore hypotheses and solutions faster so they can test many ideas without frustration

Principle · All principles
Evidence · 19 claims (10 for, 9 mixed) · 13 studies (3 causal, 3 review, 2 quant-synthesis, 2 qualitative, 2 design, 1 associational), q1–q4 · 4 of 13 report an effect size · 18 claims rest on one study

Conditional relationship

This page owns the speed of the test cycle in exploration: what a computer tool adds when it returns the consequence of a learner's idea in seconds instead of hours. The tool might be a simulation, a microworld, a real-time data logger that graphs a measurement as it is taken, a spreadsheet model, or a music or drawing program that plays back a change at once. For a learner investigating how a system behaves (a principle or causal model in science or economics, or the effect of a choice in music or art), such a tool is expected to do three things. It lets them try more hypotheses, methods and strategies in the same time. It keeps a failed idea cheap, so they keep going rather than giving up in frustration. And it frees class time for interpreting the results. The expected payoff is better conceptual-understanding of the system and better inquiry skill: forming a testable hypothesis, changing one thing at a time, setting an achievable goal. This holds only when three conditions are met: the exploration is guided (a question or prediction for each round, prompts or a person who responds), the learner can already work the tool, and the round ends by connecting what was found to the canonical idea.

No claim in this wiki tests the speed of the cycle itself, that is, the same exploration with fast against slow feedback and learning measured. The nearest evidence comes in three parts:

So the page's own relationship is a design proposal resting on a plausible mechanism (more and cheaper cycles) and on one confounded course comparison. The guidance condition is the best-evidenced part. A fast tool used for unguided exploration is the configuration the evidence most clearly expects to fail.

How the converted siblings sit beside this one:

  • Inquiry-Based Learning owns the general relationship: investigating a question with guidance against being told or left alone. This page is the tool layer inside it.
  • Simulations & Immersive Virtual Environments owns practising a real, risky or costly task in a designed representation, with a debrief. This page is about exploring a problem space to understand it, where the sheer number of tries is the point.
  • Digital Learning owns the argument that a medium is not a method. Here the method is the guided hypothesis–test cycle, and the tool only makes it faster.
  • Scaffolding and Fading owns how support during the rounds is raised and withdrawn.
  • In the Cognitive Apprenticeship pattern, exploration is the last phase, after modelling, coaching and articulation. This page supplies that phase's tool.

Default design, while the relationship is untested

The earlier page's guidance, made concrete. Each step is a proposal; its evidence status is stated beside it. The doses are untested proposals, to be set locally and recorded.

  1. Choose a tool whose answer comes back in seconds and whose variables the learner can change one at a time. Examples: a simulation or microworld with sliders, a sensor that graphs as it measures, a spreadsheet model with input cells, a sequencer or drawing program with undo. Check that the variables the learner can change are the ones the target idea is about. Evidence: a real-time tool halved one lab's duration (Real-time data collection made the momentum lab substantially more efficient, cutting lab duration from two hours to one [+W]). Choosing variables to match the target is an untested proposal.
  2. Give learners time with the tool before the content task (5–10 minutes, longer for an unfamiliar interface). Set a no-stakes task: find what each control does, produce one given result. Evidence: teachers found students needed time to learn and explore a technology before gaining mathematics understanding from it (Teachers reported that students needed dedicated time to learn and explore technologies before being expected to gain mathematics understanding from them [~W]). A well-designed simulation may need very little: one sixth grader began interacting with an energy simulation within 10 seconds and started making sense of it without instructions (A student began interacting with the sim within 10 seconds and verbalized sense-making without explicit guidance [+W], one interview). The duration is an untested proposal.
  3. Open every round with a question and a written prediction. Before touching the tool, the learner writes what they will change, what they expect and why (one line each). Evidence: guidance during inquiry raised outcomes and inquiry-skill use (Guidance during inquiry learning improves post-inquiry learning outcomes (d = 0.50) [+M]; Guidance during inquiry learning produces a medium-to-large improvement in learners' use of inquiry skills during the inquiry [+M]). Prediction-first as the form of guidance is an untested proposal; the meta-analysis found type of guidance did not moderate the effect on learning outcomes or on inquiry-skill use.
  4. Guide during the round; do not leave novices to explore freely. Use a prompt sheet or built-in prompts ("What did you hold constant?", "Does the result match your prediction?"). A teacher or tutor who sees a learner changing several things at once asks them to change one. Evidence: Unassisted discovery produces less learning than explicit instruction, while discovery enhanced with guidance outperforms other instruction [~S]; Supplementing simulation-based inquiry with learner guidance has a modest effect on learning outcomes (d = 0.49) and inquiry skills (d = 0.41), independent of guidance type (review attribution to D'Angelo et al., 2014) [+M]. Guidance also raised what learners achieved during the inquiry (Guidance during inquiry learning improves performance success during the inquiry (d = 0.71), with type of guidance moderating this effect [+M]).
  5. Ask for several cycles per round, logged. In a 15–20 minute round, expect three to five hypothesis–test cycles. Each goes in a log of four columns: change, prediction, result, explanation. A learner who runs many unlogged trials is clicking, not testing (see the diagnosis table below). The number of cycles is an untested proposal from the earlier page's "test many ideas without frustration".
  6. Make the worksheet ask for reasoning, not values to copy. Ask for a prediction, an explanation of a mismatch, or a rule that covers all the results, never a table filled from the screen. Evidence: students using a collision-carts simulation said the worksheet "simply presents a set of values to copy" and generated little thinking (Some students found the guided-inquiry worksheet activities insufficiently thought-provoking and the data exchange troublesome [~W]).
  7. Spend the time the tool saves on interpretation. Use the time no longer spent collecting data for comparing groups' results, arguing about anomalies, and building or revising a model. Do not use it to cover more material. Evidence: the authors of the real-time data study argue this is where their gains came from (The authors of one physics-course study attribute its technology-supported gains to the student-centred, active-engagement pedagogy the technology was used within [~W]; Students in a technology-enhanced modeling physics class learned significantly more than students in an otherwise identical no-technology class [+W]). Both rest on one confounded study.
  8. Close each round by connecting results to the canonical idea, then check without the tool. Give 10 minutes: groups state their rule, the teacher (or a worked answer) gives the canonical explanation and names where groups' rules fit or fail, and each learner answers one new question unaided, away from the tool. Evidence: an immediate advantage of simulation labs over physical labs did not persist to the midterm and final (AI-generated and prebuilt simulations produced higher immediate conceptual-assessment scores than physical laboratory work, but the advantage did not persist in later examinations [~M]), and simulation games support near transfer more robustly than far or long-term transfer (The review reports that simulation games support near transfer of knowledge but evidence for long-term or far transfer is less robust [~M]). The closing routine is an untested proposal.
  9. Hand over goal-setting across sessions. The earlier page's aim was that learners learn to set achievable goals and form and test hypotheses themselves. A default arc of four to six sessions:
  10. Sessions 1–2: the teacher sets the question and fixes which variables may change; a prompt sheet runs throughout.
  11. Sessions 3–4: the learner chooses the variable and writes the hypothesis from a template ("If I change __, then __ will __, because __"); prompts only on request.
  12. Sessions 5 onward: the learner sets their own achievable goal and plans a sequence of tests. The teacher reviews the plan in two minutes before the round, and the learner explores with no prompt sheet.
  13. Move a learner to the next stage when, on two consecutive rounds, they write a testable hypothesis, change one thing at a time, and explain the result unaided. Keep a learner who already does this unaided at the later stage (The advantage of guidance begins to recede only when learners have sufficiently high prior knowledge to provide internal guidance [~M]).

The arc and its thresholds are untested proposals; the handing over of support follows Scaffolding and Fading.

When a condition is not met: if no tool returns answers quickly, use the same guided cycle with a slower method and fewer, better-chosen tests. Speed is a convenience the guidance does not depend on. If no one can guide (self-paced use), build the question, the prediction box, the prompts and the closing unaided question into the tool or the worksheet. If the tool cannot vary what the target idea is about, do not use it for that objective.

Fitting the design to a situation

The three facts that most change the decision:

  • What the learner already knows about the system and the tool. A novice on the target relation needs a set question, fixed variables and prompts; an advanced learner can set their own goals and may find prompts redundant; anyone new to the interface needs time with it first. If unstated, ask: can they predict one effect in this system now, and can they work the tool's controls?
  • Who can see the exploration as it happens. Guidance is the best-evidenced condition, but a teacher with thirty screens sees few of them. If unstated, ask: during a round, who will notice a learner changing everything at once or clicking without predicting, and what can they do about it?
  • What must the learner be able to do afterwards, and without the tool? The options are an explanation of the system, an inquiry skill, a product made with the tool, or a later exam taken without it. If unstated, ask: what will the learner do at the end, with which aids, and when?
If the brief says… Then change… Basis
Learners are novice on the target relation (e.g. a first encounter with how a variable affects a system) Set the question and fix which variables may change; require a written prediction before each test; keep a prompt sheet throughout; end with the canonical explanation Unassisted discovery produces less learning than explicit instruction, while discovery enhanced with guidance outperforms other instruction [~S]
Learners are advanced, or already design fair tests unaided Let them set their own goal and test plan; drop the prompt sheet; review the plan briefly instead The advantage of guidance begins to recede only when learners have sufficiently high prior knowledge to provide internal guidance [~M] (a review's statement; gives no threshold)
Learners are child (primary age) Fewer variables (two or three), shorter rounds (10 minutes), the teacher reading predictions aloud with the group; allow a prediction to be drawn or pointed at rather than written Supplementing simulation-based inquiry with learner guidance has a modest effect on learning outcomes (d = 0.49) and inquiry skills (d = 0.41), independent of guidance type (review attribution to D'Angelo et al., 2014) [~M] reports that age moderates the effect but not how; the specific changes are untested
Learners are adults in a work setting (e.g. exploring a pricing model, a process simulation or a spreadsheet of their own operation) Start from a question they bring from their work; use their own data or model where possible; let them set the goal from session 1, with a facilitator reviewing the test plan untested proposal
The learner is new to the interface, or has low confidence with computers Add a 10-minute no-stakes familiarisation task before the content question; pair a confident and a less confident user, swapping the controls each round Teachers reported that students needed dedicated time to learn and explore technologies before being expected to gain mathematics understanding from them [~W] (teacher reflection, one unit)
Learners have low literacy, or work in a second language Replace the written log with a picture or screenshot log (before, after, thumbs up or down for the prediction); give the hypothesis template orally and in the home language where possible untested proposal
classroom, one teacher and many learners at screens Run rounds in pairs at one device; groups hold up or post their prediction before testing so the teacher can scan for one-change-at-a-time; the teacher circulates to groups whose log is empty untested proposal; guidance evidence is pooled across settings and does not separate a teacher circulating from built-in prompts (Guidance during inquiry learning produces a medium-to-large improvement in learners' use of inquiry skills during the inquiry [~M])
Mixed mode: some learners remote, some in the room, or some on phones only Use a browser tool that works on a phone, or have remote learners send predictions to an in-room partner who runs the tool and shares the screen; collect every prediction in one shared place (chat, shared document) before each test untested proposal
Setting is online-self-paced, no teacher present Build the question, prediction box, prompts and closing unaided question into the tool or worksheet; show the canonical explanation only after the learner has submitted a rule Supplementing simulation-based inquiry with learner guidance has a modest effect on learning outcomes (d = 0.49) and inquiry skills (d = 0.41), independent of guidance type (review attribution to D'Angelo et al., 2014) [+M] (effects did not depend on guidance type, which suggests built-in prompts can serve; untested as such)
Scarce facilitator time (a volunteer, a busy shift, a tutor checking in briefly) Spend the facilitator's minutes on the prediction before the round and the rule after it, not on watching the round; a printed prompt card covers the middle untested proposal
Single session only, or a fixed short time budget (e.g. 50 minutes) 5 min tool familiarisation, 5 min question and predictions, 20 min exploration with the log (three to five cycles), 10 min groups state their rule and hear the canonical explanation, 5 min one unaided question, 5 min spare; claim only immediate understanding untested proposal; immediate gains may not last (AI-generated and prebuilt simulations produced higher immediate conceptual-assessment scores than physical laboratory work, but the advantage did not persist in later examinations [~M])
The outcome that counts is a later exam taken without the tool, or high stakes Add an unaided question at the end of every round and a spaced return to the idea in later weeks; judge readiness only on unaided answers, never on what was achieved with the tool AI-generated and prebuilt simulations produced higher immediate conceptual-assessment scores than physical laboratory work, but the advantage did not persist in later examinations [~M]; The review reports that simulation games support near transfer of knowledge but evidence for long-term or far transfer is less robust [~M]
The brief names its own routine (its own worksheet, rubric, lab report template or check-in) Keep it; change only what the evidence says matters: add a prediction line before each test if it lacks one, and replace any copy-the-values item with a reasoning item Some students found the guided-inquiry worksheet activities insufficiently thought-provoking and the data exchange troublesome [~W]
The brief's own assessment rewards what was achieved with the tool Run it as given; add one unaided question beside it, not instead of it untested proposal
The brief uses its own categories or criteria for a good result (a rubric, a list of accepted explanations) Use those categories in the prediction template and the closing rule; do not substitute this page's untested proposal
The product is for a real audience (a composition, a design, a model presented to a client or the public) Keep exploration and the product separate: rounds of trial on drafts, then a final piece in which the learner says which tried options they chose and why; the audience sees the product, the reasoning log stays with the learner and teacher untested proposal
A whole-task performance the brief wants kept whole (e.g. tuning a whole system, composing a whole piece) Explore inside the whole task: each round changes one feature of the whole, never a detached part; log the change against the whole result untested proposal
The material is private or sensitive (learners' own finances, health or workplace data in a model) Let the learner choose to use their own data or a provided dataset; the log stays with the learner and is shared only by their choice untested proposal
Learners report an error they actually made (e.g. changing two variables at once and crediting the wrong one) Replay it as a compared case: show both runs, ask which variable caused the change, then run the one-change version untested proposal
No fast tool is available Run the same guided cycle with physical materials or a paper model; plan fewer, better-chosen tests, and use the prediction log to make each one count untested proposal; the guidance evidence is not specific to computer tools (Guidance during inquiry learning improves post-inquiry learning outcomes (d = 0.50) [+M])

Ten of the twenty rows cite a claim. None of them tests a fast tool against a slow one. Most carry guidance evidence pooled across inquiry and simulation studies into the row's setting, and the age, interface and worksheet rows rest on reviews or single reports that say so. The rest are untested proposals.

Observation, state and explanation

What can be observed in exploration is the learner's actions: what they changed, how many things at once, whether a prediction came first, how many cycles they ran, what they wrote, and what they said when the result did not match. "Understands the system", "is testing hypotheses" and "is frustrated" are inferred states. Results achieved with the tool show what the learner can find with it, not what they can explain or predict without it.

Response Plausible explanations Observation that could change the interpretation
Many rapid trials, no predictions, an empty log Playing with the tool rather than testing; does not know how to form a hypothesis; the question is too open Stop and ask for one prediction for the next change; if they cannot give one, the question needs narrowing or a modelled example of a prediction.
Changes several variables at once and credits the wrong one Does not yet hold the one-change-at-a-time strategy; the interface makes several changes easy; they are aiming for a target result rather than an explanation Ask which change caused the result; set a task where two simultaneous changes would give an ambiguous result and see whether they notice.
Accurate rule with the tool, cannot answer an unaided question The rule is tied to the screen; the unaided question differs more than intended; the canonical idea was not connected Give a matched unaided question with a diagram of the tool's display, then one without; ask them to explain the rule in their own words.
Stops after one or two cycles, says "it doesn't work" Frustration (the condition the earlier page wanted fast tools to prevent); the goal is unachievable; trouble with the tool Ask what they were trying to get it to do; offer a smaller goal; watch them use one control to separate a tool problem from a reasoning problem.
Says it was fun and they "figured it out", but the log shows no stated rule Enjoyment without a stated conclusion; the conclusion was formed but not written; low literacy Ask them to tell or show the rule; compare with the canonical idea.

These rows are proposals, untested with learners as diagnostics. A probe is itself guidance and changes what it measures, so record each one.

Evidence and qualifications

No claim in the wiki compares fast with slow test cycles on learning, or varies the number of cycles. The claims below test neighbouring parts of the relationship.

Learners, tools and outcomes differ across these claims (undergraduate physics cohorts, pooled inquiry studies, primary-school children, a single interview), so do not rank them by effect label. Only the guidance claims compare conditions with any control. None measures delayed retention of what was learned through exploration.

The evidence by question

Further evidence, not yet read against this model

Objective and learner-valued goal

The designer's objective is usually an explanation of how the system behaves, and the inquiry skill of forming and testing hypotheses, both shown without the tool at a stated horizon. Elicit separately what the learner values. Students in one simulation lesson valued the fun, figuring things out themselves, and "thinking like real scientists" (Students reported that the virtual-lab lesson made learning fun, helped them figure out concepts themselves, and supported thinking like real scientists [+W]). That matches the earlier page's hope that learners gain "a sense of what it is like to be a scientist". It can also diverge from the objective. A learner who values making something happen on the screen may run many satisfying trials and state no rule. A learner who values getting the right answer may want the canonical explanation before exploring at all. Ask, record both, and do not read enjoyment as understanding.

For example, a music student exploring a sequencer may want to make a track that sounds good to friends, while the designer's objective is understanding how rhythm subdivisions combine. Rounds that vary one subdivision at a time serve the objective. A final track the student chooses, with a note on which tried options they kept, serves the learner's goal.

What would revise this model?

The expectation should weaken if studies comparing the same guided exploration with fast and slow feedback find no difference in learning. It should also weaken if more cycles per round add nothing once guidance is held constant, or if the time saved by a tool goes into covering more material with no gain. If guided exploration with a fast tool still loses to explicit instruction for novices on aligned conceptual tests, the tool belongs after instruction, as practice, not as the way in. If built-in prompts work as well as a responding teacher (as the guidance-type finding hints), the classroom rows can lean on the tool rather than on circulating. Do not protect the model by calling every failure "not enough guidance" or "not enough cycles" after the fact.

Speed of the test cycle, number of cycles, guidance, immediate understanding, delayed retention and inquiry skill are separate claims. The present evidence establishes that guidance matters and that tools can shorten a lab. It establishes no number of cycles, no familiarisation time and no hand-over criterion.

Claims

Examples

Key Sources

  • Collins, Allan. (1988). Cognitive Apprenticeship and Instructional Technology. Technical Report. BBN Report No. 6899, BBN Labs, Inc. https://eric.ed.gov/?id=ED331465
  • Brewe, E., Sawtelle, V., and Pamela, P. (2007). Impacts of real-time data collection on introductory algebra-based physics. https://doi.org/10.48550/arxiv.0709.2738
Page metadata
Field Value
type principle
id exploration-tools-fast-hypothesis-testing
aliases use-lab-technology-to-free-time-for-analysis
title Use computer tools that let students explore hypotheses and solutions faster so they can test many ideas without frustration
description For learners exploring how a system behaves, a computer tool that returns the result of each idea in seconds is expected to let them test more hypotheses with less frustration and spend the saved time interpreting, improving conceptual understanding and inquiry skill when the exploration is guided; no claim in the wiki tests the speed of the cycle itself, though guidance during inquiry is supported.
status review
canonical True
generated by claude/unspecified on 2026-10-15
trust unverified — no one has confirmed this page's content
sources 2 — collins-1988, brewe-2007

Frontmatter is the page's machine-readable record — see the Schema & Guide.